fix(adapters): retain late chat tool-call index aliases

[devswha]

Summary

• Keep a streamed Chat tool call intact when the provider introduces it by ID, later sends that ID with an index, and then continues with index-only argument fragments. Previously the final fragment created a second unnamed call, split the JSON arguments, and failed the response.
• Retain the observed index as an alias of the existing pending call while preserving its stable translator-budget key. Interleaved calls, existing index precedence, ID fallback, and byte limits keep their existing behavior.
• Extend the existing stream-assembly regression and document the supported sequence in the adapter reference.

Verification

• The original fixture failed through both the adapter and the Responses JSON bridge; an index-present-first control completed successfully.
• bun test tests/adapters/openai/openai-chat-parallel-stream.test.ts tests/adapters/openai/openai-chat-hardening.test.ts tests/adapters/openai/openai-chat-eof.test.ts — 124 passed, 0 failed. Coverage includes interleaved calls, ambiguous missing associations, existing ID/index collision behavior, stable budget ownership, and byte-limit boundaries.
• bun run typecheck — passed on the final head.
• cd docs-site && bun run build — passed, 425 pages.
• bun run privacy:scan — passed.
• bun run test:changed on final head c8240c51d — 6,152 passed, 4 skipped, 0 failed across 334 files.
• CodeRabbit manual review completed with no actionable comments.
• Draft remains pending the full local-CI readiness gate and maintainer approval to run fork CI.
• A separate full-suite review baseline completed with 19,250 passed, 16 skipped, 9 default-timeout failures, and 1 related asynchronous error. This branch does not claim a green full-suite run.

Checklist

• Scope stays focused and avoids unrelated cleanup.
• Docs or release notes were updated when needed.
• Security-sensitive changes were reviewed for secrets, auth, and unsafe defaults. No authentication, credential, or destination behavior changes.

Review readiness

• All CI tests are green on my local testing.

• I pushed my PR to the latest dev commit.

• I resolved all correct Codex and CodeRabbit findings.

• My PR is ready for review.

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: ASSERTIVE

Plan: Team

Run ID: 74d03833-0b24-4b61-8034-af11a173db7a

📥 Commits

Reviewing files that changed from the base of the PR and between 6585e6a and c8240c5.

📒 Files selected for processing (3)

• docs-site/src/content/docs/reference/adapters.md
• src/adapters/openai-chat.ts
• tests/adapters/openai/openai-chat-parallel-stream.test.ts

Included review availability: Your plan provides up to 10 included reviews per hour; 6 remain after this review.

---

📝 Walkthrough

Walkthrough

The OpenAI Chat adapter now retains index aliases for streamed tool calls. Tests cover late indexes, index-only fragments, conflicting or duplicate IDs, and argument-budget limits. Documentation describes the identity-preservation behavior.

Changes

OpenAI Chat tool-call identity

Layer / File(s)	Summary
Resolve streamed tool-call identities src/adapters/openai-chat.ts	parseStream stores the first observed index alias and uses it to match later fragments without changing the key that owns the argument budget.
Validate identity and budget behavior tests/adapters/openai/openai-chat-parallel-stream.test.ts, docs-site/src/content/docs/reference/adapters.md	Tests cover late indexes, ambiguous index-only fragments, conflicting and duplicate IDs, alias retention, and argument-byte limits. Documentation describes the streaming behavior.

Estimated code review effort: 3 (Moderate) | ~20 minutes

Merge Risk: ⚪ Minimal · up to c8240

The OpenAI Chat adapter now correctly assembles streamed tool calls when IDs and indexes arrive in different fragments, with regression coverage for parallel calls and budget limits. No merge-blocking risk is identified.

Suggested reviewers: lidge-j

🚥 Pre-merge checks | ✅ 4 | ❌ 1

❌ Failed checks (1 warning)

Check name	Status	Explanation	Resolution
Docstring Coverage	⚠️ Warning	Docstring coverage is 0.00% which is insufficient. The required threshold is 80.00%. Docstring coverage is scoped to functions touched by this diff. Analyzed 2 functions across 2 files. (1 skipped: 1 …	Write docstrings for the functions missing them to satisfy the coverage threshold.

✅ Passed checks (4 passed)

Check name	Status	Explanation
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title clearly and concisely describes the main change: retaining late index aliases for streamed chat tool calls in the adapters.

Full details: Docstring Coverage

Explanation

Docstring coverage is 0.00% which is insufficient. The required threshold is 80.00%. Docstring coverage is scoped to functions touched by this diff. Analyzed 2 functions across 2 files. (1 skipped: 1 unsupported.)

• Fix all pre-merge checks with AI

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}

[github-actions]

✅ Deterministic PR hygiene checks passed.

[github-actions]

⏳ DRAFT

• review readiness checklist open (2/4 boxes ticked).

What to do

• Tick all four boxes in the PR description once you're done (currently 2/4).

Review readiness checklist

• ⬜ All CI tests are green on my local testing.
• ✅ I pushed my PR to the latest dev commit.
• ✅ I resolved all correct Codex and CodeRabbit findings.
• ⬜ My PR is ready for review.

2/4 boxes ticked.

This PR stays in draft until every box above is ticked.

Hygiene

✅ Deterministic PR hygiene checks passed.

[devswha]

@coderabbitai review

[coderabbitai]

✅ Action performed

Review finished.

Note: CodeRabbit is an incremental review system and does not re-review already reviewed commits. This command is applicable only when automatic reviews are paused.

[lidge-jun]

리뷰 · 우선순위 68 / 80

이 PR은 Chat Completions 스트림에서 도구 호출이 먼저 ID만, 다음에 ID+index, 그다음 index만 있는 argument 조각으로 이어질 때, 마지막 조각이 이름 없는 두 번째 호출을 만들어 JSON이 쪼개지던 버그를 고칩니다. src/adapters/openai-chat.ts의 PendingToolCall에 indexKey 별칭을 두고, 예산(budget) 소유 키는 처음 연 키를 유지한 채 나중에 관측된 index만 붙입니다. 이미 다른 index가 묶인 ID를 다시 다른 index로 옮기지는 않습니다. 병렬 호출·기존 index 우선·ID fallback·바이트 한도는 테스트로 고정했습니다.

지금 dev의 조립 로직은 rawIndex가 있으면 무조건 i:N 키로 새 슬롯을 찾기 때문에, ID-only로 시작된 호출과 이후 index-only 조각이 만나지 못합니다. OpenRouter 등 일부 공급자 스트림이 이 순서를 씁니다. 문서 reference/adapters.md에 지원 시퀀스를 한 단락 추가한 것도 운영자에게 도움이 됩니다. Draft이고 test:changed는 rebase 후 기록 예정이라고 본문에 있습니다.

테스트가 “index만으로 서로 다른 unindexed 두 호출을 추측하지 않는다”, “충돌 ID에서 index가 이긴다”, “중복 ID first-match”, “늦은 index에서도 argument byte limit”까지 넓혀서, 별칭 추가가 예전의 보수적 실패 모드를 무너뜨리지 않는지 잘 막아 두었습니다. types/config 분할과 무관합니다.

src/adapters/openai-chat.ts PendingToolCall.indexKey - ID-first 호출에 늦은 index 별칭; budget key 불변
같은 파일 find 경로 - indexKey로 재검색; 첫 index만 bind
openai-chat-parallel-stream.test.ts - T9b 확장·interleave·no-guess·limit 9/10
docs adapters.md - ID→index→index-only 조립 설명
Draft - ready 전 test:changed 결과 기입 필요

메인테이너의 판단이 필요한 지점

• 이 별칭 동작을 openai-chat 공통 계약으로 바로 받을지
• Draft 해제 조건으로 parallel-stream + hardening + eof 그린만으로 충분한지

너의 추천
스트림 도구 호출 실실패 수정으로 merge 쪽. 테스트 그린 확인 후 Draft 해제·merge.

이 댓글은 grok-bot이 작성했습니다

[lidge-jun]

Integrated into dev through #3702 (merge eeca697), preserving the original Hako-authored change and adding index/identity validation fixes. Exact-head Linux and macOS CI passed; the pinned remote macOS full suite passed 19,551 tests with 15 skips. Closing the original after verifying dev ancestry.